Constituent ontologists would seem to have a serious problem accounting for accidental change. Suppose an avocado goes from unripe to ripe over a two day period. That counts as an accidental change: one and the same substance (the avocado) alters in respect of the accidental property of being unripe. It has become different qualitatively while remaining the same numerically.
This is a problem for constituent ontologists if C-ontologists are committed to what Michael J. Loux calls "Constituent Essentialism." ("What is Constituent Ontology?" Metaphysics: Aristotelian, Scholastic, Analytic, Ontos Verlag 2012, Novak et al. eds., p. 52) Undoubtedly, many of them are, if not all. Constituent Essentialism is the C-ontological analog of mereological essentialism. We can put it like this:
Constituent Essentialism: A thing has each of its ontological parts necessarily. This implies that a thing cannot gain or lose an ontological part without ceasing to be same
Mereological Essentialism: A thing has each of its commonsense parts necessarily.
This implies that a thing cannot gain or lose a commonsense part without ceasing
to be the same thing.
To illustrate, suppose an ordinary particular (OP) such as our avocado is a bundle of compresent universals. The universals are the ontological parts of the OP as a whole. The first of the two principles entails that ordinary particulars cannot change. For accidental (alterational as opposed to existential) change is change in respect of properties under preservation of numerical diachronic identity. But preservation of identity is not possible on Constituent Essentialism. The simple bundle-of-universals theory is incompatible with the fact of change. But of course there are other types of C-ontology.
I agree with Loux that Constituent Essentialism is a "framework principle" (p. 52) of C-ontology. It cannot be abandoned without abandoning C-ontology. If an item (of whatever category) has ontological parts at all, then it is difficult to see how it could fail to have each and all of these parts essentially. And of course the fact of accidental change and what it entails, namely, persistence of the same thing over time, cannot be denied. So the 'argument from change' does seem to score against primitive versions of the bundle-of-universals theory.
I don't want to discuss whether more sophisticated C-ontological theories such as Hector Castaneda's Guise Theory escape this objection. I want to consider whether relational ontology does any better. I take relational ontology to imply that no item of any category has ontological parts. Thus R-ontology implies that no type of particular has ontological parts. A particular is just an unrepeatable. My cat Max is a particular and so are each of his material parts, and their material parts. If Max's blackness is an accident of him as substance, then this accident is a particular. The Armstrongian state of affairs of Max's being black is a particular. Mathematical sets are particulars. Particulars need not be concrete. Sets are abstract particulars in one sense of 'abstract.' Tropes are abstract particulars in another sense of 'abstract.' If an entity is not a particular, an unrepeatable, then it is a universal, a repeatable.
My question is whether we can explain real (as opposed to 'Cambridge') accidental change without positing particulars having ontological constituents. I will argue that we cannot, and that therefore R-ontology is untenable.
Lukas Novak presents an argument to the conclusion that the fact of accidental change requires the positing of particulars that have ontological constituents. Here is my take on Novak's argument:
Peter goes from being ignorant of the theorem of Pythagoras to being knowledgeable about it. This is an accidental change: one and the same concrete particular, Peter, has different properties at different times. Now a necessary condition of accidental change is that one and the same item have different properties at different times. But is it a sufficient condition? Suppose Peter is F at time t and not F at time t* (t* later than t). Suppose that F-ness is a universal but not a constituent of Peter and that Peter is F by exemplifying F-ness. Universals so construed are transcendent in the sense that they are not denizens of the world of space and time. They belong in a realm apart and are related, if they are related, to spatiotemporal particulars by the external relation of exemplification.
It follows on these assumptions that if Peter undergoes real accidental change that Peter goes from exemplifying the transcendent universal F-ness at t to not exemplifying it at t*. That is: he stands in the exemplification relation to F-ness at t, but ceases so to stand to t*. But there has to be more to the change than this. For, as Novak points out, the change is in Peter. It is intrinsic to him and cannot consist merely in a change in a relation to a universal in a realm apart. After all, transcendent universals do not undergo real change. Any change in such a universal is 'merely Cambridge' as we say in the trade. In other words, the change in F-ness when it 'goes' from being exemplified by Peter to not being exemplified by Peter is not a real change in the universal but a merely relational change. The real change in this situation must therefore be in or at Peter. For a real, not merely Cambridge, change has taken place.
Thus it seems to Novak and to me that, even if there are transcendent universals and ordinary concrete particulars, we need another category of entity to account for accidental change, a category that that I will call that of property-exemplifications. (We could also call them accidents. But we must not, pace Novak, call them tropes.) Thus Peter's being cold at t is a property-exemplification and so is Peter's not being cold at t*. Peter's change in respect of temperature involves Peter as the diachronically persisting substratum of the change, the universal coldness, and two property-exemplifications, Peter's being cold at t and Peter's being not cold at t*.
These property-exemplifications, however, are particulars, not universals even though each has a universal as a constituent. This is a special case of what Armstrong calls the Victory of Particularity: the result of a particular exemplifying a universal is a particular. Moreover, these items have natures or essences: it is essential to Peter's being cold that it have coldness as a constituent. (Thus Constituent Essentialism holds for these items. ) Hence property- exemplifications are particulars, but not bare particulars. They are not bare because they have natures or essences. Further, these property-exemplifications are abstract particulars in that they do not exhaust the whole concrete reality of Peter at a time. Thus Peter is not merely cold at a time, but has other properties besides.
It seems that the argument shows that there have to be these abstract particulars -- we could call them accidents instead of property-exemplifications -- if we are to account for real accidental change. But these partculars have constituents. Peter's coldness, for example, has Peter and coldness as constituents. It is a complex, not a simple. (If it were a simple, there would be nothing about it to tie it necessarily to Peter. Tropes are simples, so accidents are not tropes.) So it seems to me that what Novak has provided us with is an argument for C-ontology, for the view that the members of at least one category of entity have ontological constituents.
Loux's argument notwithstanding, a version of C-ontology seems to be required if we are to make sense of accidental change.
But how are accidents such as Peter's coldness connected or tied -- to avoid the word 'related' -- to a substance such as Peter?
First of all, an accident A of a substance S does not stand in an external relation to S -- otherwise a Bradleyan regress arises. (Exercise for the reader: prove it.)
Second, A is not identical to S. Peter's coldness is not identical to Peter. For there is more to Peter than his being cold. So what we need is a tie or connection that is less intimate than identity but more intimate than an external relation. The part-whole tie seems to fit the bill. A proper part of a whole is not identical to the whole, but it is not externally related to it either inasmuch as wholes depend for their identity and existence on their parts.
Can we say that Peter's accidents are ontological parts of Peter? No. This would put the cart before the horse. Peter's coldness is identity- and existence-dependent on Peter. Peter is ontologically prior to his accidents. No whole, however, is ontologically prior to its parts: wholes are identity and existence-dependent on their parts. So the accidents of a substance are not ontological parts of it. But they have ontological parts. Strangely enough, if A is an accident of substance S, then S is an ontological part of A. Substances are ontological parts of their accidents! Brentano came to a view like this.
More on Brentano later. For now, my thesis is just that the fact of real accidental change requires the positing of particulars that have ontological constituents and that, in consequence, R-ontology is to be rejected. Constituent ontology vindicatus est.